Contact details

Dr Paul P Aylin

Clinical Reader in Epidemiology & Public Health
School of Public Health

Tel: +44 (0)20 7594 3334
Email:

Dr Paul Aylin

I am a Clinical Reader in Epidemiology and Public Health and the Assistant Director of the Dr Foster Unit at Imperial.  I trained in Public Health Medicine and spent three years at the Office for National Statistics as a Medical Statistician before coming to Imperial in 1997. 

Initially I worked within the Small Area Health Statistics Unit, looking into environmental effects on health using routinely collected health data. My health services research started with work commissioned by the Bristol Royal Infirmary Inquiry, looking into mortality in paediatric cardiac surgery in Bristol and across England using routinely collected NHS data. I acted as expert witness and our analysis was cited many times in the Inquiry final report, and indeed our estimates of “excess” deaths were widely quoted elsewhere. The Inquiry itself represented a turning point in the medical establishment in the UK, and much of the work we have done since arose out of the consequences of the Inquiry. We published our account of the work in the Lancet. (Aylin P, Alves B, Best N, Cook A, Elliott P, Evans SJ, Lawrence AE, Murray GD, Pollock J, Spiegelhalter D. Comparison of UK paediatric cardiac surgical performance by analysis of routinely collected data 1984-96: was Bristol an outlier? Lancet 2001 Jul 21;358(9277): 181-187.)

My next major contribution was for the Shipman Inquiry, an inquiry into the serial killing GP. We were commissioned by the Inquiry to determine and develop appropriate statistical methods for monitoring over time, mortality at GP and practice level. I again acted as an expert witness for the Shipman Inquiry and as part of a team, developed our methodology and made recommendations to the Inquiry. The results of our work were incorporated in the fifth Inquiry report in 2004. (www.the-shipman-inquiry.org.uk. Reports, The 5th Report, Chapter 14--The Monitoring of Mortality Rates among the Patients of General Practioners.) We again published our work in the Lancet (Aylin P, Best NG, Bottle A and Marshall C. Following Shipman:  a pilot system for monitoring mortality rates in primary care. The Lancet 2003;362:485-91).

My work as Assistant Director of the Dr Foster Unit has allowed me to apply methods similar to those we developed for the Shipman Inquiry to provide near real-time monitoring of secondary care.  It is used by both managers and clinicians in over two thirds of acute trusts to monitor quality of care.  As well as highlighting outcomes such as high mortality, it highlights low mortality and other outcomes such as length of stay, day case rates, and readmissions.  Publication of our analyses has led to documented reductions in mortality (Jarman B, Bottle A, Aylin P, Browne M. Monitoring changes in hospital standardised mortality ratios. BMJ 2005, 330: 115) and (Wright J, Dugdale B, Hammond I, Jarman B, Neary M et al. Learning from death: a hospital mortality reduction programme. JR Soc Med, June 2006, 99:303-308.) The Dr Foster Unit at Imperial College London is funded by a grant from Dr Foster Intelligence (an independent health service organisation).

Current work

Work on predicting high impact users of emergency inpatient care has resulted in a tool used by both PCTs and GPs to focus more community orientated care on those patients identified (Bottle A, Aylin P, Majeed A. Identifying patients at high risk of emergency hospital admissions:  a logistics regression analysis. JR Soc Med, Aug 2006; 99:406-414).

We are carrying out work to validate our risk prediction models used within the RTM. We are comparing risk prediction models for mortality based on HES with published results based on data derived from three national clinical databases: the National Cardiac Surgical Database, the National Vascular Database and the Colorectal Cancer Study (Aylin P, Bottle A, Majeed A. Use of Administrative data or clinical databases as predictors of risk of death in hospital: comparison of models. BMJ 2007;334: 1044).

We are looking into adapting patient safety indicators from work originally commissioned by the Agency for Healthcare Research and Quality in the US. This will result in a number of indicators that could be derived from existing HES and NWCS data, and could be built into existing tools, or result in a entirely standalone patient safety tool.

Having obtained the ONS mortality field linked to HES data, we have been investigating its use for examining long term survival rates for several conditions and procedures. Initially this has focussed on CABGs and MIs, but could be extended to more conditions. However, it is unlikely it could be used for looking at cancer survival rates because of the difficulty in determining the date of diagnosis.

We have carried out a comparison of HES data with the UK Colorectal Cancer Database compiled by the Association of Coloproctology of Great Britain and Ireland. This compares both counts and mortality for seven procedures, and has been carried out in conjunction with the ACPGBI. A similar study comparing Hospital Episode Statistics with the Vascular Society of Great Britain and Ireland’s National Vascular Database has also been carried out in conjunction with the Vascular Society (2007. Descriptive study comparing routine hospital administrative data with the Vascular Society of Great Britain and Ireland's National Vascular Database. Eur J Vasc Endovasc Surg. 33: 461-465).

Future research

We have a strong and comprehensive programme for future research, a summary of which is given below. Our longer term goals are to become a leading centre for health services research, particularly in the utilisation of routinely collected data. We would like to focus on primary care data sets. We also intend to explore international data sets.

NIHR Research Centres for NHS Patient Safety & Service Quality
I have also been successful in a joint bid with St. Mary’s and the Hammersmith Hospitals Trust for an NIHR Patient Safety Centre. My particular programme of research will look into the monitoring of patient safety indicators, and how those indicators are used to improve patient care.

The utility of primary care data in benchmarking and monitoring outcomes
Hospital Episode Statistics are a rich source of information about the experience patients have of secondary care. However, secondary care is just the tip of the illness iceberg, and people consult their GPs much more frequently. We intend to work with the Dept Primary Care and Social Medicine on primary care data taken from information systems in North West Thames and other areas as available to examine the utility of these data for benchmarking and performance monitoring as well as in the management of individual patients.  This will be dependent on obtaining the data.

ONS linked mortality file
We need to determine the utility of the HES-ONS linked field for measuring longer term mortality and survival. This is not as straightforward as using the current outcome measure of 30 day in-hospital deaths. To calculate survival data for cancer diagnoses for example, the admission where the cancer is first diagnosed (i.e. the anniversary date) needs to be determined. This requires linkage of records back to individual patients. A proxy for this might be to focus on definitive curative procedures, for example mastectomy, although there is no guarantee that this would be the first procedure carried out. The result of this work could result in a tool which provides one year cancer survival by centre, and survival from other procedures (e.g. CABG).

Improving case mix adjustment models
A criticism of both clinicians and academics is that HES data do not contain enough data to sufficiently adjust for case-mix. Provisional work within the unit suggests that in some cases the case-mix models we use are as good as, if not better, than some case-mix models based on clinical data. This work needs expanding, looking at other diagnoses and procedures and other outcomes.

Comparisons of HES data to clinical data sets
We have been establishing collaborative work with a number of professional bodies including the Vascular Society of Great Britain and Ireland, the Association of Coloproctology of Great Britain and Ireland and the Society of Cardiothoracic Surgeons. We would like to compare information derived from the HES data set with that which the societies themselves collect in their clinical datasets. Further collaborative work to identify and correct inconsistencies between HES and clinical datasets might be a useful output of this work and could enhance the credibility of both sources of data. This research, if appropriate, will be designed to assist in the development of systems for linking clinical and administrative data sets and comparing data sets in a routine manner.

Analyses available at patient level, to derive incidence data
Further investigation of patient linked data could result in a mapping tool that not only provided standardised admission rates, but more importantly incidence rates by area.

Quantifying false alarm rates and successful detection rates
Within the RTM, we give rough guides to the performance of each CUSUM chart, i.e. the ability to successfully detect a problem of there is one, and the false alarm rate. These performance measures require further investigation, so that appropriate thresholds can be set for processes that appear to signal frequently such as length of stay and some specific diagnoses groups such as HSMRS which generate many alerts (both positive and negative). These revised chart performance measures need to be made available within the RTM and some commentary on their revision needs to be published in peer reviewed papers.

Patient safety analyses using HES
Patient safety is increasingly important to healthcare agencies. We will consult and collaborate with experts in the field on determining new outcome measures that might reflect aspects of patient safety. This might include the incorporation of intensive care data to determine “near misses” and other process data such as blood transfusion rates to measure good practice. This work might result in a completely new adverse event or patient safety tool. Findings from this work will be published in an academic journal.

Investigation of length of stay as a continuous variable
Length of stay is a difficult parameter to model as average length of stay can be grossly distorted by a few long staying residents on a hospital ward. Currently we use fairly crude measures such as median length of stay, or treat length of stay as a binary outcome (e.g. patients staying less than or more than a certain period). There has been some work carried out which allows more flexible modelling of length of stay that would allow us to produce better estimates of bed days saved, excess bed days etc. This work needs to be applied to the RTM.

The use of different statistical process control charts
Currently, we use individual risk adjusted log-likelihood CUSUM charts for all process within the RTM. There may be diagnoses, procedures and outcomes where this is not the most appropriate method and better methods might be employed. In addition, the current CUSUM charts are useful for looking at individual processes, but are not so useful in terms of monitoring process over a large number of units. If we are to pursue a tool for Strategic Health Authorities to be able to use our analyses for clinical governance, we need to look at other ways of analysing and presenting these results. Spiegelhalter has already suggested a number of methods for this purpose which the Healthcare Commission currently use. We need to implement our own versions of these charts but with the same flexibility of analysis afforded by the RTM web front end.

Investigation of further disease and specialty specific outcomes
The HES database is a rich source of information, and we would like to explore outcomes other than mortality and readmission rates for their usefulness in monitoring healthcare performance.

Teaching

I have taught medical students right from the beginning of my medical career as a junior doctor. After commencing training in public health medicine, I helped to run weekly educational seminars and revision courses for the professional exams. I then moved to a job within the Office of National Statistics as a medical statistician. While there, as part of remit to improve the teaching of death certification to medical students and doctors, I helped develop a teaching pack incorporating a video and tutorial material. I also attempted to evaluate the video through a randomised controlled trial. It was after my move to Imperial College as a Clinical Senior Lecturer in 1997 that I had more formal teaching responsibilities. I taught epidemiology and public health to groups of medical students attached to our department for a month at a time. This was repeated for 10 months of the year. Shortly after this we revised the course for the new curriculum to include several modules throughout the six years.

I was departmental head of undergraduate teaching until 2005, when I took on the role of Divisional Head of Undergraduate Teaching Quality. I am now divisional director of postgraduate taught courses.

In addition I developed and ran a module on the Social Medicine BSc called Epidemiology for Clinicians and up until 2005, jointly ran the introduction to epidemiology and statistics course on the Modern Epidemiology MSc. I have now developed a module within the MPH course on health information.

Other teaching commitments
Year 1.  Medical undergraduate course - Foundations of Evidence Based Medicine Course
Lecturer

PBL Medical undergraduate course
Development of Problem Based Learning cases and tutor

Year 6. Medical undergraduate course - Evidence for Clinical Practice 
Course leader. Completely redesigned this year.

MSc in Modern Epidemiology
Lecturer in “introduction to epidemiology” module

MPH
Module organiser within the MPH course for health information.

Guest lectures
Guest lecturer on variety of external courses including a session on Child Health at University of Roehampton and a lecture on Sources of Health Data for a City University MSc course

External examiner
For Sheffield and Nottingham BSc courses

PhD and MD supervision

Previous teaching contributions
Health and Environment MSc (now ceased)
Lecturer within Health and Environment option

PhD examiner
Two occasions for the London School of Hygiene and Tropical Medicine in December 2003 and for Sheffield University in October 2004

2nd year medical undergraduate Public Health course (now ceased)
Development of lectures, course material and tutorial work

Social Medicine BSc
Epidemiology for Clinicians module (now ceased)
Module leader
 

Selected Publications